- Email: [email protected]
Membrane biology: Do glycolipid microdomains really exist?
R114
Dispatch
Membrane biology: Do glycolipid microdomains really exist? Nigel M. Hooper
Glycolipid membrane domains have been suggested to have a number of physiological functions, but do they actually exist in vivo or are they artefacts of extraction procedures? Recent data go some way towards showing that such glycolipid domains really are present within both model and cellular membranes. Address: School of Biochemistry and Molecular Biology, University of Leeds, Leeds LS2 9JT, UK. Current Biology 1998, 8:R114–R116 http://biomednet.com/elecref/09609822008R0114 © Current Biology Ltd ISSN 0960-9822
The outer leaflet of the mammalian cell membrane is enriched in sphingolipids — glycosphingolipids and sphingomyelin — which associate laterally with one another, probably through weak interactions between the carbohydrate heads of the glycosphingolipids [1]. The headgroups of the sphingolipids occupy larger areas in the plane of the outer leaflet of the lipid bilayer than their predominantly saturated acyl chains, with the consequent spaces between the acyl chains being filled by cholesterol molecules (Figure 1). The nature of the phospholipids occupying the cytoplasmic leaflet of these cholesterol/ glycosphingolipid-rich domains is unknown. Depleting cells of cholesterol with sterol-binding agents, such as filipin, nystatin or methyl-β-cyclodextrin, or of glycosphingolipids with fumonisin B1 has revealed that such lipid
domains are involved in a diverse array of biological processes, including clathrin-independent endocytosis (potocytosis), transcytosis, cholesterol transport, signal transduction and bacterial and viral internalisation [2,3]. Cholesterol/glycosphingolipid-rich domains are insoluble at 4°C in certain non-ionic detergents, such as Triton X100, so a common means of identifying proteins and lipids that associate with them is to assess their detergentinsolubility. Because of their high lipid content, these detergent-insoluble complexes float to a low density during sucrose density gradient centrifugation [4–6]. The resulting low-density, detergent-insoluble membrane fraction — termed DIGs for detergent-insoluble glycolipidenriched domains [7] — is enriched not only in cholesterol and glycosphingolipids, but also in specific proteins. These include glycosyl-phosphatidylinositol (GPI)-anchored proteins, which are localised to the outer leaflet of the bilayer, certain transmembrane proteins, such as the haemagglutinin protein of influenza virus and intestinal epithelial sucrase-isomaltase, and several acylated (myristoylated and/or palmitoylated) proteins, such as the Src-family tyrosine kinases and heterotrimeric G proteins, which are localised to the cytoplasmic leaflet (Figure 1) [8]. A crucial question has been raised by these studies, however, and that is: do domains rich in clustered GPIanchored proteins, glycosphingolipids, cholesterol and
Figure 1 A model for the organisation of lipid domains and caveolae in the plasma membrane. The glycolipid-rich domains (red) segregate from the other regions (grey) of the bilayer, possibly forming an annulus around the neck of the caveolae. Individual lipids and proteins may move between these domains and other regions of the bilayer, including the caveolae. The lipid bilayer in glycolipid-rich domains is asymmetric, with sphingomyelin (orange) and glycosphingolipids (GSLs; red) enriched in the outer leaflet. Cholesterol (dark green) is present in both leaflets and fills the spaces under the headgroups of the sphingolipids. GPI-anchored proteins (blue) are attached to the outer leaflet of the bilayer, and acylated proteins — such as Src-family kinases (purple) — are integrated into the cytoplasmic leaflet. Some polypeptide-anchored proteins — such as influenza virus haemagglutinin (yellow) — are also associated with the glycolipid-rich domains. Modified from [1,7].
Dispatch
acylated proteins actually exist in the cell membrane? Or does the extraction of membranes with detergent cause the more hydrophobic acyl chains of the GPI anchors, glycosphingolipids and acylated proteins to cluster artefactually together with cholesterol as the phospholipids, which generally have less hydrophobic acyl chains, are selectively removed? Recently, more biophysical approaches have been employed in an attempt to answer this question. In model membranes, physiological concentrations of cholesterol and sphingolipid induce formation of a liquidordered phase, which has properties that are intermediate between those of the fluid and gel phases. The liquidordered phase is characterised by tight acyl chain packing and relatively extended acyl chains. Liquid-ordered phase bilayers are insoluble in Triton X-100. Model membranes with a similar lipid composition to that of cellular DIGs were also found to be insoluble in Triton X-100, as was the GPI-anchored alkaline phosphatase inserted into such model membranes [9], suggesting that the lipids themselves determine the detergent insolubility of both the lipids and proteins in DIGs. But these observations left open the possibility that the detergent extraction used to isolate DIGs might alter the composition of the domains or induce their formation from previously uniform lipid mixtures. In an attempt to resolve this conundrum, a fluorescence quenching assay has been used to detect phase separations in the presence of cholesterol [10]. A liquidordered phase was seen to form in model membranes with similar lipid, particularly sphingomyelin, compositions to the plasma membrane. The presence of cholesterol, again at a concentration similar to that found in the plasma membrane, promoted the formation of the liquid-ordered phase at 37°C. Critically, the detergentinsolubility of cholesterol-containing model membranes closely correlated with the amount of liquid-ordered phase, as detected by fluorescence quenching with nitroxide-labelled lipids. This led Ahmed et al. [10] to conclude that the detergent-insoluble membranes isolated from cells are likely to exist in the liquid-ordered phase before detergent extraction, and that one of the more important roles of cholesterol and sphingolipids in cell membranes may be to induce formation of the liquidordered phase. In a separate approach, Jacobson and coworkers [11] have used single-particle tracking to follow the movements of two components of DIGs — the GPI-anchored Thy-1 protein and the glycosphingolipid GM1 — on the surface of fibroblasts. Single-particle tracking allows the measurement with nanometer precision of the movement on the surface of a cell of individual molecules specifically labelled with colloidal gold or fluorescent particles.
R115
Figure 2
A model for the molecular associations of a GPI-anchored protein in a cell membrane. (a) GPI-anchored proteins (blue) are transiently associated with and confined by glycosphingolipid domains (red). (b) Slow, anomalous diffusion occurs when molecules encounter regions rich in proteins (yellow) in the membrane. (c) Fast Brownian motion occurs in relatively unobstructed regions of the membrane. Modified from [11].
Video-enhanced bright-field microscopy was used to record the movement of membrane components over a defined period of time. In this way, the movements of Thy-1 and GM1 were categorised into four modes of lateral transport (Figure 2). First, fast diffusion due to unobstructed Brownian motion within the lipid bilayer. Second, slow, anomalous diffusion equivalent to movement through protein-rich domains. Third, diffusion confined to 325–370 nm diameter regions. And fourth, a fraction of the molecules was essentially stationary on the 6.6 second time scale of the experiment. Longer observations (60 seconds) showed that Thy-1 and GM1 are both transiently confined for 7–9 seconds to regions averaging 260–370 nm in diameter. About 36% of Thy-1 and GM1 undergo this confined diffusion. In contrast, only 16% of fluorescein phosphatidylethanolamine — a phospholipid analogue not expected to be found in DIGs — displayed confined diffusion. Furthermore, the phosphatidylethanolamine was confined to regions averaging approximately 230 nm in diameter, significantly (1.5fold) smaller than those observed for Thy-1 and GM1. Reducing the glycosphingolipid expression of the cells by about 40% with a glucosylceramide synthase inhibitor caused the percentage of trajectories exhibiting confinement, and the size of the confining domain for the GPIanchored Thy-1, to be reduced approximately 1.5-fold. Interestingly, extraction of the cells with Triton X-100 left the fraction of molecules confined, and the domain sizes for Thy-1 and GM1, unchanged. This result is consistent with the preferential association of GPI-anchored proteins with glycosphingolipids in distinct domains in the cell membrane. The results also suggest that the confining domains may be the in vivo equivalent of the detergentinsoluble membrane fractions.
R116
Current Biology, Vol 8 No 4
Interestingly, single particle tracking studies of the GPIanchored and transmembrane isoforms of the human neural cell adhesion molecule (NCAM) in fibroblasts and muscle cells revealed that about 30% of both isoforms experienced transient confinement for approximately 8 seconds within regions of approximately 300 nm diameter [12]. Thus, the same mechanism of confinement appears to apply to both the GPI-anchored and polypeptide-anchored forms of NCAM. Diffusion of the protein within the regions of approximately 300 nm diameter was anomalous, consistent with movement through a dense field of obstacles, which was likened to the movement of a ball in a pinball machine. Simson et al. [12] concluded that the membrane appears as a mosaic, containing regions that permit free diffusion as well as regions in which NCAM is transiently confined to small or more extended domains. So there is now more direct evidence to indicate that the components of DIGs do indeed exist in discrete domains in the cell membrane in vivo. Several questions still remain, however. Are the GPI-anchored Thy-1 and the glycosphingolipid GM1 co-localised to the same confining domains? From the observations with NCAM, are all or just a subset of polypeptide-anchored proteins similarly confined as GPI-anchored proteins? And are acylated proteins, such as the Src-family kinases, localised to the cytoplasmic leaflet of domains in which GPI-anchored proteins are localised to the outer leaflet? Simultaneously tracking the movement of two different molecules with fluorescent particles having different colours [11] may go some way towards answering these questions. Furthermore, what is the function, or functions, of such lipid domains, and how are they related to caveolae, the flask-shaped invaginations of the plasma membrane that appear to be involved in clathrin-independent endocytosis and to act as signal transduction centres? Experiments using cationic colloidal silica particles to coat the surface of endothelial cells in situ before the isolation of caveolae have indicated that GPI/cholesterol/glycosphingolipid-rich domains do not completely equate with caveolae [13,14]. From these studies it appears that the GPI/cholesterol/ glycosphingolipid-rich domains form an annulus around the neck of the caveolae (Figure 1), so the two are physically closely associated and probably functionally linked. As caveolae themselves are enriched in cholesterol and glycosphingolipids, co-isolate with GPI/cholesterol/glycosphingolipid-rich domains on sucrose density gradient centrifugation, and are affected when cells are depleted of cholesterol or glycosphingolipids, it is often difficult to ascertain which of these structures is actually involved in a particular process. Further techniques need to be developed before it will be possible routinely to dissect out these two structurally similar lipid domains in living cells, so that their respective biological functions can be
investigated. However, the clustering of GPI-anchored proteins and certain other membrane proteins, cholesterol and glycosphingolipids in distinct domains within membranes undoubtedly facilitates molecular interactions by bringing molecules into close physical association and/or by increasing the local concentration of specific molecules that are required for specialised cellular functions. Acknowledgements I thank Ed Parkin for critically reading this manuscript.
References 1. Simons K, Ikonen E: Functional rafts in cell membranes. Nature 1997, 387:569-572. 2. Baorto DM, Gao Z, Malaviya R, Dustin ML, van der Merwe A, Lublin DM, Abraham SN: Survival of FimH-expressing enterobacteria in macrophages relies on glycolipid traffic. Nature 1997, 389:636-639. 3. Stevens VL, Tang J: Fumonisin B1-induced sphingolipid depletion inhibits vitamin uptake via the glycosylphosphatidylinositolanchored folate receptor. J Biol Chem 1997, 272:18020-18025. 4. Brown DA, Rose JK: Sorting of GPI-anchored proteins to glycolipid-enriched membrane subdomains during transport to the apical cell surface. Cell 1992, 68:533-544. 5. Chang W-J, Ying Y, Rothberg KG, Hooper NM, Turner AJ, Gambliel HA, De Gunzburg J, Mumby SM, Gilman AG, Anderson RGW: Purification and characterisation of smooth muscle cell caveolae. J Cell Biol 1994, 126:127-138. 6. Lisanti MP, Scherer PE, Viugirene J, Tang Z, Hermanowski-Vosatka A, Tu Y-H, Cook RF, Sargiacomo M: Characterisation of caveolin-rich membrane domains isolated from an endothelial-rich source: implications for human disease. J Cell Biol 1994, 126:111-126. 7. Parton RG, Simons K: Digging into caveolae. Science 1995, 269:1398-1399. 8. Harder T, Simons K: Caveolae, DIGs, and the dynamics of sphingolipid-cholesterol microdomains. Curr Opin Cell Biol 1997, 9:534-542. 9. Schroeder R, London E, Brown D: Interactions between saturated acyl chains confer detergent resistance on lipids and glycosylphosphatidylinositol (GPI)-anchored proteins: GPIanchored proteins in liposomes and cells show similar behaviour. Proc Natl Acad Sci USA 1994, 91:12130-12134. 10. Ahmed SN, Brown DA, London E: On the origin of sphingolipid/cholesterol-rich detergent-insoluble cell membranes: Physiological concentrations of cholesterol and sphingolipid induce formation of a detergent-insoluble, liquidordered lipid phase in model membranes. Biochemistry 1997, 36:10944-10953. 11. Sheets ED, Lee GM, Simson R, Jacobson K: Transient confinement of a glycosylphosphatidylinositol-anchored protein in the plasma membrane. Biochemistry 1997, 36:12449-12458. 12. Simson R, Yang B, Moore SE, Doherty P, Walsh FS, Jacobson KA: Structural mosaicism on the submicron scale in the plasma membrane. Biophys J 1998, in press. 13. Schnitzer JE, McIntosh DP, Dvorak AM, Liu J, Oh P: Separation of caveolae from associated microdomains of GPI-anchored proteins. Science 1995, 269:1435-1439. 14. Liu J, Oh P, Horner T, Rogers RA, Schnitzer JE: Organised endothelial cell surface signal transduction in caveolae distinct from glycosylphosphatidylinositol-anchored protein microdomains. J Biol Chem 1997, 272:7211-7222.
Dispatch
Membrane biology: Do glycolipid microdomains really exist? Nigel M. Hooper
Glycolipid membrane domains have been suggested to have a number of physiological functions, but do they actually exist in vivo or are they artefacts of extraction procedures? Recent data go some way towards showing that such glycolipid domains really are present within both model and cellular membranes. Address: School of Biochemistry and Molecular Biology, University of Leeds, Leeds LS2 9JT, UK. Current Biology 1998, 8:R114–R116 http://biomednet.com/elecref/09609822008R0114 © Current Biology Ltd ISSN 0960-9822
The outer leaflet of the mammalian cell membrane is enriched in sphingolipids — glycosphingolipids and sphingomyelin — which associate laterally with one another, probably through weak interactions between the carbohydrate heads of the glycosphingolipids [1]. The headgroups of the sphingolipids occupy larger areas in the plane of the outer leaflet of the lipid bilayer than their predominantly saturated acyl chains, with the consequent spaces between the acyl chains being filled by cholesterol molecules (Figure 1). The nature of the phospholipids occupying the cytoplasmic leaflet of these cholesterol/ glycosphingolipid-rich domains is unknown. Depleting cells of cholesterol with sterol-binding agents, such as filipin, nystatin or methyl-β-cyclodextrin, or of glycosphingolipids with fumonisin B1 has revealed that such lipid
domains are involved in a diverse array of biological processes, including clathrin-independent endocytosis (potocytosis), transcytosis, cholesterol transport, signal transduction and bacterial and viral internalisation [2,3]. Cholesterol/glycosphingolipid-rich domains are insoluble at 4°C in certain non-ionic detergents, such as Triton X100, so a common means of identifying proteins and lipids that associate with them is to assess their detergentinsolubility. Because of their high lipid content, these detergent-insoluble complexes float to a low density during sucrose density gradient centrifugation [4–6]. The resulting low-density, detergent-insoluble membrane fraction — termed DIGs for detergent-insoluble glycolipidenriched domains [7] — is enriched not only in cholesterol and glycosphingolipids, but also in specific proteins. These include glycosyl-phosphatidylinositol (GPI)-anchored proteins, which are localised to the outer leaflet of the bilayer, certain transmembrane proteins, such as the haemagglutinin protein of influenza virus and intestinal epithelial sucrase-isomaltase, and several acylated (myristoylated and/or palmitoylated) proteins, such as the Src-family tyrosine kinases and heterotrimeric G proteins, which are localised to the cytoplasmic leaflet (Figure 1) [8]. A crucial question has been raised by these studies, however, and that is: do domains rich in clustered GPIanchored proteins, glycosphingolipids, cholesterol and
Figure 1 A model for the organisation of lipid domains and caveolae in the plasma membrane. The glycolipid-rich domains (red) segregate from the other regions (grey) of the bilayer, possibly forming an annulus around the neck of the caveolae. Individual lipids and proteins may move between these domains and other regions of the bilayer, including the caveolae. The lipid bilayer in glycolipid-rich domains is asymmetric, with sphingomyelin (orange) and glycosphingolipids (GSLs; red) enriched in the outer leaflet. Cholesterol (dark green) is present in both leaflets and fills the spaces under the headgroups of the sphingolipids. GPI-anchored proteins (blue) are attached to the outer leaflet of the bilayer, and acylated proteins — such as Src-family kinases (purple) — are integrated into the cytoplasmic leaflet. Some polypeptide-anchored proteins — such as influenza virus haemagglutinin (yellow) — are also associated with the glycolipid-rich domains. Modified from [1,7].
Dispatch
acylated proteins actually exist in the cell membrane? Or does the extraction of membranes with detergent cause the more hydrophobic acyl chains of the GPI anchors, glycosphingolipids and acylated proteins to cluster artefactually together with cholesterol as the phospholipids, which generally have less hydrophobic acyl chains, are selectively removed? Recently, more biophysical approaches have been employed in an attempt to answer this question. In model membranes, physiological concentrations of cholesterol and sphingolipid induce formation of a liquidordered phase, which has properties that are intermediate between those of the fluid and gel phases. The liquidordered phase is characterised by tight acyl chain packing and relatively extended acyl chains. Liquid-ordered phase bilayers are insoluble in Triton X-100. Model membranes with a similar lipid composition to that of cellular DIGs were also found to be insoluble in Triton X-100, as was the GPI-anchored alkaline phosphatase inserted into such model membranes [9], suggesting that the lipids themselves determine the detergent insolubility of both the lipids and proteins in DIGs. But these observations left open the possibility that the detergent extraction used to isolate DIGs might alter the composition of the domains or induce their formation from previously uniform lipid mixtures. In an attempt to resolve this conundrum, a fluorescence quenching assay has been used to detect phase separations in the presence of cholesterol [10]. A liquidordered phase was seen to form in model membranes with similar lipid, particularly sphingomyelin, compositions to the plasma membrane. The presence of cholesterol, again at a concentration similar to that found in the plasma membrane, promoted the formation of the liquid-ordered phase at 37°C. Critically, the detergentinsolubility of cholesterol-containing model membranes closely correlated with the amount of liquid-ordered phase, as detected by fluorescence quenching with nitroxide-labelled lipids. This led Ahmed et al. [10] to conclude that the detergent-insoluble membranes isolated from cells are likely to exist in the liquid-ordered phase before detergent extraction, and that one of the more important roles of cholesterol and sphingolipids in cell membranes may be to induce formation of the liquidordered phase. In a separate approach, Jacobson and coworkers [11] have used single-particle tracking to follow the movements of two components of DIGs — the GPI-anchored Thy-1 protein and the glycosphingolipid GM1 — on the surface of fibroblasts. Single-particle tracking allows the measurement with nanometer precision of the movement on the surface of a cell of individual molecules specifically labelled with colloidal gold or fluorescent particles.
R115
Figure 2
A model for the molecular associations of a GPI-anchored protein in a cell membrane. (a) GPI-anchored proteins (blue) are transiently associated with and confined by glycosphingolipid domains (red). (b) Slow, anomalous diffusion occurs when molecules encounter regions rich in proteins (yellow) in the membrane. (c) Fast Brownian motion occurs in relatively unobstructed regions of the membrane. Modified from [11].
Video-enhanced bright-field microscopy was used to record the movement of membrane components over a defined period of time. In this way, the movements of Thy-1 and GM1 were categorised into four modes of lateral transport (Figure 2). First, fast diffusion due to unobstructed Brownian motion within the lipid bilayer. Second, slow, anomalous diffusion equivalent to movement through protein-rich domains. Third, diffusion confined to 325–370 nm diameter regions. And fourth, a fraction of the molecules was essentially stationary on the 6.6 second time scale of the experiment. Longer observations (60 seconds) showed that Thy-1 and GM1 are both transiently confined for 7–9 seconds to regions averaging 260–370 nm in diameter. About 36% of Thy-1 and GM1 undergo this confined diffusion. In contrast, only 16% of fluorescein phosphatidylethanolamine — a phospholipid analogue not expected to be found in DIGs — displayed confined diffusion. Furthermore, the phosphatidylethanolamine was confined to regions averaging approximately 230 nm in diameter, significantly (1.5fold) smaller than those observed for Thy-1 and GM1. Reducing the glycosphingolipid expression of the cells by about 40% with a glucosylceramide synthase inhibitor caused the percentage of trajectories exhibiting confinement, and the size of the confining domain for the GPIanchored Thy-1, to be reduced approximately 1.5-fold. Interestingly, extraction of the cells with Triton X-100 left the fraction of molecules confined, and the domain sizes for Thy-1 and GM1, unchanged. This result is consistent with the preferential association of GPI-anchored proteins with glycosphingolipids in distinct domains in the cell membrane. The results also suggest that the confining domains may be the in vivo equivalent of the detergentinsoluble membrane fractions.
R116
Current Biology, Vol 8 No 4
Interestingly, single particle tracking studies of the GPIanchored and transmembrane isoforms of the human neural cell adhesion molecule (NCAM) in fibroblasts and muscle cells revealed that about 30% of both isoforms experienced transient confinement for approximately 8 seconds within regions of approximately 300 nm diameter [12]. Thus, the same mechanism of confinement appears to apply to both the GPI-anchored and polypeptide-anchored forms of NCAM. Diffusion of the protein within the regions of approximately 300 nm diameter was anomalous, consistent with movement through a dense field of obstacles, which was likened to the movement of a ball in a pinball machine. Simson et al. [12] concluded that the membrane appears as a mosaic, containing regions that permit free diffusion as well as regions in which NCAM is transiently confined to small or more extended domains. So there is now more direct evidence to indicate that the components of DIGs do indeed exist in discrete domains in the cell membrane in vivo. Several questions still remain, however. Are the GPI-anchored Thy-1 and the glycosphingolipid GM1 co-localised to the same confining domains? From the observations with NCAM, are all or just a subset of polypeptide-anchored proteins similarly confined as GPI-anchored proteins? And are acylated proteins, such as the Src-family kinases, localised to the cytoplasmic leaflet of domains in which GPI-anchored proteins are localised to the outer leaflet? Simultaneously tracking the movement of two different molecules with fluorescent particles having different colours [11] may go some way towards answering these questions. Furthermore, what is the function, or functions, of such lipid domains, and how are they related to caveolae, the flask-shaped invaginations of the plasma membrane that appear to be involved in clathrin-independent endocytosis and to act as signal transduction centres? Experiments using cationic colloidal silica particles to coat the surface of endothelial cells in situ before the isolation of caveolae have indicated that GPI/cholesterol/glycosphingolipid-rich domains do not completely equate with caveolae [13,14]. From these studies it appears that the GPI/cholesterol/ glycosphingolipid-rich domains form an annulus around the neck of the caveolae (Figure 1), so the two are physically closely associated and probably functionally linked. As caveolae themselves are enriched in cholesterol and glycosphingolipids, co-isolate with GPI/cholesterol/glycosphingolipid-rich domains on sucrose density gradient centrifugation, and are affected when cells are depleted of cholesterol or glycosphingolipids, it is often difficult to ascertain which of these structures is actually involved in a particular process. Further techniques need to be developed before it will be possible routinely to dissect out these two structurally similar lipid domains in living cells, so that their respective biological functions can be
investigated. However, the clustering of GPI-anchored proteins and certain other membrane proteins, cholesterol and glycosphingolipids in distinct domains within membranes undoubtedly facilitates molecular interactions by bringing molecules into close physical association and/or by increasing the local concentration of specific molecules that are required for specialised cellular functions. Acknowledgements I thank Ed Parkin for critically reading this manuscript.
References 1. Simons K, Ikonen E: Functional rafts in cell membranes. Nature 1997, 387:569-572. 2. Baorto DM, Gao Z, Malaviya R, Dustin ML, van der Merwe A, Lublin DM, Abraham SN: Survival of FimH-expressing enterobacteria in macrophages relies on glycolipid traffic. Nature 1997, 389:636-639. 3. Stevens VL, Tang J: Fumonisin B1-induced sphingolipid depletion inhibits vitamin uptake via the glycosylphosphatidylinositolanchored folate receptor. J Biol Chem 1997, 272:18020-18025. 4. Brown DA, Rose JK: Sorting of GPI-anchored proteins to glycolipid-enriched membrane subdomains during transport to the apical cell surface. Cell 1992, 68:533-544. 5. Chang W-J, Ying Y, Rothberg KG, Hooper NM, Turner AJ, Gambliel HA, De Gunzburg J, Mumby SM, Gilman AG, Anderson RGW: Purification and characterisation of smooth muscle cell caveolae. J Cell Biol 1994, 126:127-138. 6. Lisanti MP, Scherer PE, Viugirene J, Tang Z, Hermanowski-Vosatka A, Tu Y-H, Cook RF, Sargiacomo M: Characterisation of caveolin-rich membrane domains isolated from an endothelial-rich source: implications for human disease. J Cell Biol 1994, 126:111-126. 7. Parton RG, Simons K: Digging into caveolae. Science 1995, 269:1398-1399. 8. Harder T, Simons K: Caveolae, DIGs, and the dynamics of sphingolipid-cholesterol microdomains. Curr Opin Cell Biol 1997, 9:534-542. 9. Schroeder R, London E, Brown D: Interactions between saturated acyl chains confer detergent resistance on lipids and glycosylphosphatidylinositol (GPI)-anchored proteins: GPIanchored proteins in liposomes and cells show similar behaviour. Proc Natl Acad Sci USA 1994, 91:12130-12134. 10. Ahmed SN, Brown DA, London E: On the origin of sphingolipid/cholesterol-rich detergent-insoluble cell membranes: Physiological concentrations of cholesterol and sphingolipid induce formation of a detergent-insoluble, liquidordered lipid phase in model membranes. Biochemistry 1997, 36:10944-10953. 11. Sheets ED, Lee GM, Simson R, Jacobson K: Transient confinement of a glycosylphosphatidylinositol-anchored protein in the plasma membrane. Biochemistry 1997, 36:12449-12458. 12. Simson R, Yang B, Moore SE, Doherty P, Walsh FS, Jacobson KA: Structural mosaicism on the submicron scale in the plasma membrane. Biophys J 1998, in press. 13. Schnitzer JE, McIntosh DP, Dvorak AM, Liu J, Oh P: Separation of caveolae from associated microdomains of GPI-anchored proteins. Science 1995, 269:1435-1439. 14. Liu J, Oh P, Horner T, Rogers RA, Schnitzer JE: Organised endothelial cell surface signal transduction in caveolae distinct from glycosylphosphatidylinositol-anchored protein microdomains. J Biol Chem 1997, 272:7211-7222.